UI Researchers Evaluate Proposed Changes to Fannie/Freddie

In June the Federal Housing Finance Agency's (FHFA) issued a proposed capital standard for the government-sponsored enterprises (GSEs), Freddie Mac and Fannie Mae.  Three Urban Institute researchers have analyzed the rule with an eye to answering two questions: how well it will align risk and capital across the various mortgage attributes and how the capital requirement might vary across the business cycle.  Requiring too much capital raises mortgage rates and reduces homeownership; too little results in insolvency and financial crisis.

The proposed rule includes two alternative leverage ratio proposals.  Under the first, the GSEs would be required to hold capital equal to 2.5 percent of total assets and off-balance sheet guarantees.  This approach would require them to hold a minimum amount of capital that does not differentiate between the risk characteristics of assets and guarantees.  The second alternative would require the GSEs to hold capital equal to 1.5 percent of trust assets and 4 percent of non-trust assets.  This approach differentiates between the greater funding risks of non-trust assets and the lower funding risks of the trust assets while increasing the capital requirements for both relative to the current statutory requirements.

Credit risk is the largest component of the proposed standard at a combined $112 billion before credit risk transfers and $90.5 billion afterward.  Even after transfers it accounts for about half the $180.9 billion in capital required as of the end of FY 2017.  That amount also includes a going-concern buffer of $39.9 billion, an operational risk charge of $4.3 billion, and a market risk component of $19.4 billion.  

Edward Golding, Laurie Goodman, and Jun Zhu say that the capital standards are more than an academic exercise.  They will likely be used to govern pricing for the duration of the GSEs' conservatorship which is now in its eleventh year.  They add that, given the deep divisions in Congress, conservatorship could continue for a long time.

To answer the first question the authors computed the capital requirements for a large variety of mortgages.  The second involved computing the capital requirements at various times over the business cycle. There is a great deal of computation and the authors make various assumptions. 

The report is extremely long and even the most simple and spare summary of the basis and conclusions is lengthy.  We have broken this into to articles, this first addresses the methodology and the alignment of the proposal with mortgage attributes. 

To determine capital requirements, FHFA took various mortgage attributes and assigned risk multipliers to each.  The categories included loan age, pay history, FICO score, property and occupancy type, debt-to-income and mark-to-market loan-to-value rations (MMLTV), loan size and purpose. These were used to calculate the base credit risk for each loan and adjust it with appropriate risk multipliers. The calculation also accounts for credit enhancement through mortgage insurance (MI) including counterparty credit risk.

The authors used a Fannie Mae database and, using the FHFA proposal, computed the net credit capital requirement for each mortgage at the end of each year from 2002 to 2016, using a different model depending on whether the mortgage is new, seasoned and performing, delinquent, or reperforming.  At any point the percentage in each bucket will vary so the GSE portfolio is analyzed at various times over the business cycle.  It would be a mistake, they say, to consider only newly originated mortgages as they tend to represent less than 10 percent of the GSEs' portfolios.

The largest risk attributes for 30-year fixed-rate mortgages are the origination LTV ratios and FICO scores and the FHFA capital requirements vary significantly by these attributes, ranging from 33 basis points for the lowest risk to as high at 1,176 for high LTV low FICO score loans with an overall average of 186 at the end of FY 2017.  To determine if these requirements are appropriate the authors compared them with stressed losses, using loans originated in 2007 to run the comparisons. The results show there needed to be 170 basis points of capital if each loan on the books today went through the 2007 experience.  Thus, on average, the capital requirements are high enough for the GSEs to have survived the Great Recession.

However, the analysis also found that, while FHFA had calculated the needed capital appropriately for the losses expected for various FICO scores and low LTV's they are overcapitalizing for low-FICO, high LTV loans, that is, overestimating potential losses relative to less risky loans. "These results stem from the fact that in a stress scenario, all mortgages perform worse, but the relative differential between mortgages with weaker credit scores and those with stronger credit scores is less than during normal periods, something the authors suggest FHFA should take into account when they revise the capital requirement framework.

Layering of factors such as a single borrower, high DTI ratios, and small loan sizes can also raise risk but FHFA has capped risk multiplies at 3 for loans with LTVs above 95 percent to encourage affordable housing, leaving the bulk of lending uncapped. In some examples of layering those uncapped risks results risk multipliers and thus base capital requirements much higher than actual losses would indicate.  The authors urge FHFA to do a more exhaustive review of the risks of layering, especially since the results could drive some populations to FHA where there is no risk-based pricing.  The penalties for layering may also raise fair lending issues.

FHFA reduces capital requirements where a third party assumes some credit risk; in the case of MI an average reduction of 37 percent.  With the 2007 loans the average severity for insured loans was 34 percent with 21 percent MI recovery.  This implies that the 37 percent reduction in capital is in line with historic MI effectiveness but does not acknowledge post-crisis industry changes including increased capital requirements for MI venders and contract changes making it more difficult to curtail insurance payouts.  Similarly, the standard does not allow enough relief for credit risk transfers.

They also find that using historical loss data and failing to acknowledge improved appraisal standards have resulted in higher capital requirements for both rate and cash-out refinances than necessary.  They also suggest basing capital requirements for delinquency on 90+ days rather than 30+ would give the capital requirement less volatility and be less likely to penalize borrowers who occasionally miss a payment, i.e. those with lower FICO scores.

Part two of this summary will summarize the UI analysts' assessment of the capital standard proposal's accommodation to the business cycle and a discussion of UI's findings.